ARTIFICIAL NEURAL NETWORK APPROACH FOR THE PREDICTION OF EFFLUENTS STREAMS FROM A WASTEWATER TREATMENT PLANT: A CASE STUDY IN KOCAELI (TURKEY)

Abstract

A three-layer Artificial Neural Network (ANN) model was employed to develop and estimate the effluent stream parameters of two different wastewater treatment plants (WWTP) in Kocaeli, Turkey. The chemical oxygen demand (COD), suspended solid (SS), pH and temperature as the output parameters were estimated by five input parameters such as flow rate, COD, pH, SS and temperature. The ANN model was developed with 400 data sets for prediction of effluent pH, temperature, COD and SS. The benchmark tests were employed to achieve an optimum network algorithm. The network model with optimum functions at hidden and output layers were applied for the forecasts of effluent streams of both WWTPs. The regression values of training, validation and test using this function were found as 0.94, 0.96 and 0.95, respectively. The optimum neuron numbers were determined according to the minimum mean square error values. ANN testing outputs revealed that the model exhibited well performance in forecasting the effluent pH, temperature, SS and COD values.